U.S. patent number 7,542,939 [Application Number 11/263,508] was granted by the patent office on 2009-06-02 for modeling financial instruments using bid and ask prices.
This patent grant is currently assigned to Penson Worldwide, Inc.. Invention is credited to Ronald Scott Boyd, Liam Cheung, Ralph Bruce Ferguson.
United States Patent |
7,542,939 |
Ferguson , et al. |
June 2, 2009 |
Modeling financial instruments using bid and ask prices
Abstract
A method for modeling an investment significant parameter of a
financial instrument, using a computer. At least one series of
historical bid prices of the financial instrument or historical ask
prices of the financial instrument is provided. A financial model
type that has at least one variable parameter is selected. The
variable parameter(s) of the selected financial model type is
initialized. The series of historical bid prices and/or historical
ask prices is applied to the initialized financial model type to
estimate the variable parameter(s). The resulting model of the
financial instrument may be used to predict future values of the
investment significant parameter of the financial instrument. These
predicted future values may be used to determine whether to perform
automated trades of the financial instrument.
Inventors: |
Ferguson; Ralph Bruce (Round
Rock, TX), Cheung; Liam (St. Lambert, CA), Boyd;
Ronald Scott (Austin, TX) |
Assignee: |
Penson Worldwide, Inc. (Dallas,
TX)
|
Family
ID: |
37997709 |
Appl.
No.: |
11/263,508 |
Filed: |
October 31, 2005 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20070100722 A1 |
May 3, 2007 |
|
Current U.S.
Class: |
705/37;
705/35 |
Current CPC
Class: |
G06Q
20/10 (20130101); G06Q 40/00 (20130101); G06Q
40/04 (20130101); G06Q 40/06 (20130101) |
Current International
Class: |
G06Q
40/00 (20060101) |
Field of
Search: |
;705/37,35 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
PCT International Search Report for PCT/US06/42158; Completed May
22, 2007; Mailed Sep. 5, 2007. cited by other .
Carol Alexander; Market Models: A Guide to Financial Data Analysis;
Chapters 3-13 (John Wiley & Sons, Ltd. 2001 Chichester, New
York, Weinheim, Brisbane, Singapore, Toronto). cited by
other.
|
Primary Examiner: Patel; Jagdish N
Attorney, Agent or Firm: RatnerPrestia
Claims
What is claimed:
1. A method for predicting at least one investment significant
parameter of a financial instrument for placing an order for the
financial instrument, using a computer, the method comprising the
steps of: a) generating a model of the at least one investment
significant parameter of the financial instrument using the
computer, the model based on a set of historical quotes of at least
one of bid prices of the financial instrument or ask prices of the
financial instrument and a selected financial model type; b)
selecting at least one of a bid stream of the bid prices of the
financial instrument or an ask stream of the ask prices of the
financial instrument based on the set of historical quotes on which
the model is based; c) applying the selected at least one of the
bid stream or the ask stream to the model; d) predicting the at
least one investment significant parameter of the financial
instrument by using the computer to operate the model based on the
at least one of the bid stream or the ask stream applied to the
model in step (c); e) storing the at least one predicted investment
significant parameter of the financial instrument in the computer;
and f) placing a buy or sell order, using the computer, for the
financial instrument based on the at least one predicted investment
significant parameter, wherein step (a) includes the steps of: a1)
selecting a financial model type to be input into the computer
based on the financial instrument, the financial model type having
at least one variable parameter; a2) inputting the selected
financial model type into the computer; a3) inputting the set of
historical quotes of at least one of bid prices or ask prices into
the selected financial model type; a4) initializing the selected
financial model type by initializing the at least one variable
parameter of the selected financial model type; a5) applying the
set of input historical quotes of the at least one of bid prices or
ask prices to the initialized financial model type to estimate the
at least one variable parameter; and a6) generating the model of
the at least one investment significant parameter of the financial
instrument based on the estimated at least one variable parameter,
wherein the generated model is verified to match historical
behavior of the financial instrument to within a predetermined
degree of accuracy.
2. The method according to claim 1, wherein: the at least one of
the bid stream or the ask stream selected in step (b) is at least
one of a real time bid stream of the bid prices of the financial
instrument or a real time ask stream of the ask prices of the
financial instrument; and step (d) includes predicting,
substantially in real time, at least one investment significant
parameter of the financial instrument based on the at least one of
the real time bid stream or the real time ask stream selected in
step (b).
3. The method according to claim 1, wherein: the at least one
investment significant parameter of the financial instrument
includes at least one of a future bid price or a future ask price;
the at least one of the bid stream or the offer stream selected in
step (b) is at least one of a real time bid stream of the bid
prices of the financial instrument or a real time offer stream of
the ask prices of the financial instrument; and step (d) includes
predicting, substantially in real time, at least one of future bid
prices of the financial instrument or future ask prices of the
financial instrument based on the at least one of the real time bid
stream or the real time offer stream selected in step (b).
4. The method according to claim 1, wherein step (a3) includes
inputting at least one complete consecutive series of: the bid
prices of the financial instrument spanning a predetermined period
of time; the ask prices of the financial instrument spanning the
predetermined period of time; a predetermined number of the bid
prices starting from a predetermined time; or the predetermined
number of the ask prices starting from the predetermined time.
5. The method according to claim 1, wherein: step (a3) includes
inputting at least one of: a historical time series of the bid
prices of the financial instrument, including corresponding bid
times; or a historical time series of the ask prices of the
financial instrument, including corresponding ask times; and step
(a5) includes using the computer to apply the at least one of the
historical time series of bid prices or the historical time series
of ask prices to the initialized financial model type to estimate
the at least one variable parameter.
6. The method according to claim 5, wherein step (c) includes: c1)
inputting the selected at least one of: the bid stream of the
financial instrument, including the bid prices and corresponding
bid times; or the ask stream of the financial instrument, including
the ask prices and corresponding ask times; and c2) applying the
selected at least one of the bid stream or the ask stream to the
model.
7. The method according to claim 6, wherein step (d) includes
predicting one or more future values of the investment significant
parameter of the financial instrument, including a predicted time
of each of the one or more future values of the investment
significant parameter.
8. The method according to claim 7, wherein the predicted time of
each future value of the investment significant parameter are
within a predetermined period of time after prediction of the
corresponding future value of the investment significant
parameter.
9. The method according to claim 6, wherein: the at least one
investment significant parameter of the financial instrument
includes at least one of a future bid price or a future ask price;
and step (d) includes predicting at least one of: one or more
future bid prices of the financial instrument including a predicted
bid time of each of the one or more future bid prices; or one or
more future ask prices of the financial instrument including a
predicted offer time of each of the one or more future ask
prices.
10. The method according to claim 9, wherein the predicted bid time
of each future bid price and the predicted offer time of each
future ask price are within a predetermined period of time after
prediction of the corresponding future bid prices or future ask
prices.
11. The method according to claim 1, wherein step (a3) includes
inputting at least one of: the bid prices of the financial
instrument, including corresponding bid sizes; or the ask prices of
the financial instrument, including corresponding ask sizes.
12. The method according to claim 11, wherein step (c) includes:
c1) inputting the selected at least one of: the bid stream of the
financial instrument, including the bid prices and corresponding
bid sizes; or the ask stream of the financial instrument, including
the ask prices and corresponding ask sizes; and c2) applying the
selected at least one of the bid stream or the ask stream to the
model.
13. The method according to claim 1, wherein step (a3) includes the
steps of: a3a) inputting a historical time series of the bid prices
of the financial instrument, including corresponding bid times, and
a historical time series of the ask prices of the financial
instrument, including corresponding ask times; a3b) calculating a
historical time series of a spread between the bid prices and the
ask prices of the financial instrument from the historical time
series of the bid prices and the historical time series of the ask
prices; a3c) identifying outlier bids in the historical time series
of the bid prices using the historical time series of the spread;
a3d)removing the identified outliers bids and corresponding bid
times from the historical time series of the bid prices; a3e)
identifying outlier asks in the historical time series of the ask
prices using the historical time series of the spread; and a3f)
removing the identified outliers asks and corresponding ask times
from the historical time series of the ask prices.
14. The method according to claim 1, wherein: step (a3) includes
the steps of: a3a) inputting a historical time series of the bid
prices of the financial instrument, including corresponding bid
times, and a historical time series of the ask prices of the
financial instrument, including corresponding ask times; and a3b)
calculating a historical time series of a spread between the bid
prices and the ask prices of the financial instrument from the
historical time series of the bid prices and the historical time
series of the ask prices; and step (a5) includes using the computer
to apply the historical time series of the bid prices, the
historical time series of the ask prices, and the historical time
series of the spread between the bid prices and the ask prices to
the initialized financial model type to estimate the at least one
variable parameter.
15. The method according to claim 14, wherein step (c) includes c1)
inputting the bid stream of the bid prices of the financial
instrument, including corresponding bid times, and the ask stream
of the ask prices of the financial instrument, including
corresponding ask times; c2) calculating a spread stream of a
spread between the bid prices and the ask prices of the financial
instrument from the bid stream and the ask stream; and c3) applying
the bid stream, the ask stream, and the spread stream to the
model.
16. The method according to claim 1, wherein step (a1) includes
selecting the financial model type to be one of: an expert system
model; a linear analytic model; a non-linear analytic model; a
chaotic model; a neural network model; a time delay neural network
model; a Markov-chain Monte Carlo model; a wavelet transformation
model; a regression model; a fractal model; a support vector
machine model; or a Bayesian model.
17. The method according to claim 1, wherein the financial
instrument is a publicly traded financial instrument.
18. The method according to claim 1, wherein the financial
instrument is one of a stock, a bond, a commodity, a currency, an
equity, a derivative, or a future.
19. The method according to claim 1, wherein step (c) includes the
steps of: c1) inputting the bid stream of the bid prices of the
financial instrument, including corresponding bid times, and the
ask stream of the ask prices of the financial instrument, including
corresponding ask times; c2) calculating a spread stream of a
spread between the bid prices and the ask prices of the financial
instrument from the bid stream and the ask stream; c3) identifying
outlier bids in the bid stream of the bid prices using the spread
stream; c4) removing the identified outliers bids and corresponding
bid times from the bid stream before applying the bid stream to the
model; c5) identifying outlier asks in the ask stream of the ask
prices using the spread stream; and c6) removing the identified
outliers asks and corresponding ask times from the ask stream
before applying the ask stream to the model.
20. The method according to claim 1, wherein step (d) includes
predicting at least one of: one or more future values of the
investment significant parameter of the financial instrument
including a confidence level of each of the one or more future
values of the investment significant parameter.
21. The method according to claim 1, wherein: the at least one
investment significant parameter of the financial instrument
includes at least one of a future bid price or a future ask price;
and step (d) includes predicting at least one of: one or more
future bid prices of the financial instrument including a bid
confidence level of each of the one or more future bid prices; or
one or more future ask prices of the financial instrument including
an offer confidence level of each of the one or more future ask
prices.
22. The method according to claim 1, further comprising the step
of: g) updating the model of at least one investment significant
parameter of the financial instrument using at least one of the bid
prices of the bid stream or the ask prices of the ask stream.
23. The method according to claim 1, wherein the investment
significant parameter includes at least one of: a future trade
price; a future bid price; a future ask price; a future spread; a
fair market value (FMV); an expected profit; a change in trade
price between two times; a change in bid price between two times; a
change in ask price between two times; a change in spread between
two times; a change in FMV between two times; a change in profit
between two times; a rate of change of trade price; a rate of
change of bid price; a rate of change of ask price; a rate of
change of spread; a rate of change of FMV; a rate of change of
profit; a prediction of winners and losers; or a buy/sell
instruction.
24. A computer readable medium adapted to instruct a general
purpose computer to predict at least one investment significant
parameter of a financial instrument, the computer readable medium
having stored computer executable program code for performing the
method steps of claim 1.
Description
FIELD OF THE INVENTION
The present invention concerns a method and system of modeling
financial instruments using bid and ask prices. In particular, this
method and system may allow for improved prediction of future bid
and ask prices of financial instruments and may be used to provide
information to make decisions for automated trading of various
financial instruments.
BACKGROUND OF THE INVENTION
A fundamental analysis strategy is the investment in stocks on the
basis of the value of the companies represented by the stocks. The
company's balance sheet, income statement, etc., are studied to
help determine the financial and market position of the company. If
the analysis of the company's historic growth and profit patterns
shows a steadily growing organization, and the research of the
organization and its markets show a company that is competent and
sound, a fundamental analysis approach may conclude that the
company should continue to grow and prosper.
On the other hand, a technical analysis strategy involves trying to
make profits based on the short-term swings of the market, such as,
for example, day traders, who try to take advantage of hourly or
daily price changes to make a profit. Slightly longer-term
technical analysis investors track stock price and trading volume
fluctuations over a period of a few days or weeks and trade on the
basis of recent trends. As opposed to fundamental analysis where
the emphasis is on the strength of the underlying corporation,
technical analysis focuses on patterns that appear on the
historical price charts of a specific stock and of the stock market
in general in order to help predict the future of that stock's
price. This strategy is based on the theory that certain patterns
of stock prices tend to repeat themselves over time.
The Internet provides a great variety of uses including the buying
and selling of financial instruments. The Internet has become a
major means by which investors and brokers can both monitor the
stock market and buy and sell stocks.
Although an investor does not need to be online to buy stocks,
Internet access may be of great value. The Internet offers
resources that are unmatched by any single print source. A wired
investor can get access to literally thousands of investment
services, publications, newsletters, and discussion groups. In this
manner an investor can quickly gather a large amount of information
about various financial instruments, including information about
companies whose stock may be of interest.
The stock market includes a number of features that affect the
stock investor. One of these features is the existence of agents to
facilitate the functioning of the market. Market makers,
specialists and Electronic Communications Networks (ECNs) make
market in stocks. Market makers are part of the National
Association of Securities Dealers market (NASD), and specialists
work on the New York Stock Exchange (NYSE) and other listed
exchanges. An ECN is an electronic board where buy and sell orders
may be posted by any investor worldwide. These agents serve a
similar function but there are a number of differences between
them.
The New York Stock Exchange (NYSE) is the oldest stock exchange in
the United States. The NYSE (as well as the Philadelphia, Chicago,
Boston, and Pacific Stock Exchanges) uses an agency auction market
system that is designed to allow the public to meet the public as
much as possible. The majority of trading volume (approximately
90%) occurs with no intervention from the specialist. The
responsibility of specialists is to make a fair and orderly market
in the issues assigned to them. They must yield to public orders,
which means that they may not trade for their own account when
there are public bids and asks better than their own. The
specialist has an affirmative obligation to eliminate imbalances of
supply and demand when they occur. Specialists are required to make
a continuous market. The exchange has strict guidelines for trading
depth and continuity that must be observed. Specialists are subject
to fines and censures if they fail to perform this function. NYSE
specialists have large capital requirements and are overseen by
Market Surveillance at the NYSE.
A specialist will typically maintain a narrow spread between offers
to buy and offers to sell. Generally, the trader will need access
to a professional's data feed before the trader can really see the
size of the spread.
There are over a thousand NYSE members (i.e., seats), of which
approximately a third are specialists. There are over 3000 common
and preferred stocks listed on the NYSE. On the average, each
specialist handles 6 issues. The very big stocks may have a
specialist devoted solely to them.
Every listed stock has one firm assigned to it on the floor. Most
stocks are also listed on regional exchanges in San Francisco,
Chicago, Philadelphia and Boston. All NYSE trading (approximately
80% of total volume) occurs at that post on the floor of the
specialist assigned to it.
The National Association of Securities Dealers Automated Quotation
system (NASDAQ) is an interdealer market represented by over 600
securities dealers trading more than 15,000 different issues. These
dealers are called market makers. Unlike the NYSE, the NASDAQ
market does not operate as an auction market. Instead, market
makers are expected to compete against each other by posting the
best quotes (best bid, i.e., best offer to buy, and best ask, i.e.,
best offer to sell).
A NASDAQ Level II quotation system shows all the bid offers, ask
offers, size of each offer (the order size), and the market makers
making the offers. The order size is simply the number of shares
the market maker is prepared to fill at that price. Since about
1985 the average person has had access to Level II quotes.
The Small Order Execution System (SOES) was implemented by NASDAQ
following the 1987 market crash. This system is intended to help
the small investor have his or her transactions executed without
allowing market makers to take advantage of the small investor. The
trader may see mention of "SOES Bandits" which is slang for people
who day-trade stocks on the NASDAQ using the SOES, scalping profits
on the spreads.
A firm can become a market maker on NASDAQ by applying to NASD. The
requirements include certain capital requirements, electronic
interfaces, and a willingness to make a two-sided market. The
trader must be there every day. If the trader doesn't post
continuous bids and asks every day the trader can be penalized and
not allowed to make a market for a month. Market makers are
regulated by the NASD, which is overseen by the SEC.
The brokerage firm can handle customer orders either as a broker or
as a dealer/principal. When the firm acts as a broker, it simply
arranges the trade between buyer and seller, and charges a
commission for its services. When the firm acts as a
dealer/principal, it's either buying for or selling from its own
account (to or from the customer), or acting as a market maker. The
customer is charged either a mark-up or a mark-down, depending on
whether they are buying or selling. The firm is disallowed from
charging both a mark-up (or mark-down) and a commission. Whether
acting as a broker or as a dealer/principal, the brokerage is
required to disclose its role in the transaction. However,
dealers/principals are not necessarily required to disclose the
amount of the mark-up or mark-down, although most do this
automatically on the confirmation as a matter of policy. Despite
its role in the transaction, the firm must be able to display that
it made every effort to obtain the best posted price. Whenever
there is a question about the execution price of a trade, it is
usually best to ask the firm to produce a Time and Sales report,
which allows the customer to compare all execution prices with the
actual execution price reported to the customer.
In NASDAQ, the public almost always trades with the dealer as a
counterparty instead of another public investor, making it nearly
impossible to buy on the bid or sell on the ask. Dealers can buy on
the bid even though the public is bidding at the same price.
Despite the requirement of making a market, in the case of market
makers as opposed to specialists, there is no one firm who has to
take responsibility if trading is not fair or orderly, as what
seemed to have happened during the crash of 1987. At that time,
many NASD firms simply stopped making markets or answering phones
until prices were less volatile.
Recently, Electronic Communication Networks (ECN) were established
in order to allow investors to trade NASDAQ listed stocks without
having to go through market makers, oftentimes resulting in better
prices for the investor. An ECN is an electronic system where buy
and sell orders may be posted by any investor worldwide, where any
investor or dealer may trade against that order. The best bid and
best ask orders from the ECN are posted in the NASDAQ system
alongside those of market makers.
If a trader wants to buy or sell a financial instrument, such as a
stock or other security, in an open market, the trader normally
trades via firms who act as agents who specialize in that
particular security. These firms stand ready to sell the trader a
security at the asking price (the "ask") . . . Or, if the trader
owns the security and would like to sell it, the agent buys the
security from the trader at the bid price (the "bid"). The bid and
the ask prices remain until a new price is set. The difference
between the current bid and the current ask is called the spread.
Financial instruments that are heavily traded tend to have very
narrow spreads (e.g., a few cents), but financial instruments that
are lightly traded may have spreads that are significant, even as
high as several dollars.
The width of the spread is indicative of the financial instrument's
liquidity. Liquidity basically measures the aggregate quantity
investors are willing to buy or sell of the financial instrument at
any time. In the stock market, market makers or specialists
(depending on the exchange) buy stocks from the public at the bid
and sell stocks to the public at the ask (called "making a market
in the stock"). At most times (unless the market is crashing, etc.)
these people stand ready to make a market in most stocks and often
in substantial quantities, thereby maintaining market liquidity.
Dealers earn profit by realizing a large part of the spread on each
transaction--they normally are not long-term investors.
Two types of online trading available to the public are: Internet
trading provided by firms that route a customer's order to a
trading desk or to a third party willing to pay for order flow; and
dedicated online services provided by firms where customer's orders
go directly to the exchange or ECN offering direct execution.
If the online investor uses the first type of online trading
discussed above, the customer's order may be gamed by a specialist
or market maker handling the order. Unfortunately, if this happens
to the customer, they may not be able to recognize that it has
happened from the minimal information typically provided in the
order confirmation. Typically, this type of customer only has
access to what's called Level I data--the best bid, the best ask,
the last trade, and the order size of each data respectively.
If the customer uses the second type of online trading discussed
above (i.e., the order goes from the firm directly to the
exchange), the customer most likely is looking at a NASDAQ Level II
screen. This screen shows all the bid offers, ask offers, the
recent trades, the size of each offer or trade, and the market
makers and ECNs making the offers.
An online trader connected to a web site that has a screen that
displays NASDAQ Level II data, may see the following information
streaming continuously on the screen: all bid offers, all ask
offers, all trades, the size of each offer or trade and the market
maker or ECN making the offer. This data may be refreshed as often
as ten times per second. Hence, many traders are continuously
analyzing the data on their screen all day. Moreover, unless the
trader has a prodigious memory and even then the information may
arrive too quickly to be fully read, much less utilized by the
trader. The more individual financial instruments monitored by the
trader, the greater the difficulty in utilizing the flood incoming
data. Thus, a lot of important information may escape notice.
Additionally, impatience at waiting for the desired trading
condition may cause the trader to make a trade at an inopportune
moment. Thus, an automated means of analyzing this wealth of
information is desirable.
The present invention involves methods of modeling financial
markets and automating trades to take advantage of this plethora of
bid and ask price data.
SUMMARY OF THE INVENTION
An exemplary embodiment of the present invention is a method and
system for modeling an investment significant parameter of a
financial instrument, using a computer. At least one series of
historical bid prices of the financial instrument or historical ask
prices of the financial instrument is provided. A financial model
type that has at least one variable parameter is selected. The
variable parameter(s) of the selected financial model type is
initialized. The series of historical bid prices and/or historical
ask prices is applied to the initialized financial model type to
estimate the variable parameter(s).
Another exemplary embodiment of the present invention is a method
and system for predicting future bid prices and/or future ask
prices of a financial instrument, using a computer. A model of at
least one of the bid prices or the ask prices of the financial
instrument based on a set of historical quotes of the bid prices
and/or ask prices is provided. At least one of a bid stream of the
bid prices of the financial instrument or an ask stream of the ask
prices of the financial instrument is selected based on which of
the bid prices and/or ask prices of the financial instrument are
modeled by the model. The selected bid stream and/or ask stream is
applied to the model. The model is operated to predict at least one
future bid price and/ or future ask price of the financial
instrument based on the applied bid and/or ask stream.
A further exemplary embodiment of the present invention is a method
and system for performing automated trades of at least one
financial instrument, using a computer. At least one future bid
price or future ask price of each of the financial instruments is
predicted. The future bid price and/or future ask price of each
financial instrument is predicted using at least one of a bid
stream of the bid prices of the corresponding financial instrument
or an ask stream of the ask prices of the corresponding financial
instrument. For each of the at least one financial instrument, the
predicted future bid and/or ask price(s) are compared to at least
one most recent bid and/or ask price of the corresponding financial
instrument to determine quote trend data of the corresponding
financial instrument. If the quote trend data of one of the
financial instruments meets a buy criterion for that financial
instrument, a buy order is automatically placed for the financial
instrument; and if the quote trend data of one of the financial
instruments meets a sell criterion for that financial instrument, a
sell order is automatically placed for the financial
instrument.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention is best understood from the following detailed
description when read in connection with the accompanying drawings.
Included in the drawing are the following figures:
FIG. 1 is a flow chart illustrating an exemplary method and system
of modeling an investment significant parameter of a financial
instrument according to the present invention.
FIG. 2 is a flow chart illustrating an exemplary method and system
of predicting future bid prices and/or future ask prices of a
financial instrument according to the present invention.
FIG. 3 is a flow chart illustrating an exemplary method and system
of performing automated trades of at least one financial instrument
according to the present invention.
DETAILED DESCRIPTION OF THE INVENTION
Exemplary embodiments of the present invention includes methods and
system of modeling financial instruments to predict future prices
of these financial instruments as part of exemplary technical
analysis investment strategies. These exemplary technical analysis
investment strategies may also include the automated placing of buy
and sell orders for a portfolio including one or more different
financial instruments. These financial instruments may include:
stocks, bonds, commodities, currencies, equities, derivatives,
and/or futures.
In technical analysis investment strategies, one important
consideration is the type of financial model to use. A large number
of different types of financial models have been created, such as:
expert system models; linear analytic models; non-linear analytic
models; chaotic models; neural network models; time delay neural
network models; Markov-chain Monte Carlo models; wavelet
transformation models; regression models; fractal models; support
vector machine models; and Bayesian models. Specific applications
of these types of financial models based on trade prices are known
to those skilled in the art. For example, Carol Alexander describes
a number of specific financial models in chapters 3-13 of MARKET
MODELS: A GUIDE TO FINANCIAL DATA ANALYSIS (John Wiley & Sons,
Ltd. (2001)).
Each model type includes at least one variable parameter that may
be used to match the model to the behavior of the financial
instrument. Potential variable parameters may include a numerical
value, a string to be optimized, a logical value, a conditional
rule, and/or a structural link. For example, in wavelet
transformation models, the wavelet transform coefficients are
variable parameters, while in expert system models and Bayesian
models the variable parameters may include likelihood values and/or
decision rules and in neural network models the links between nodes
may used as variable parameters.
Additionally, each model type may present a number of advantages
and a number of disadvantages when used to model the behavior of a
financial instrument. For example, neural network models may take
less time to build as compared to many other models, but because
these models are essentially black boxes and not causal. Thus, if a
neural network model diverges, it may be difficult to identify this
divergence quickly as the exact process used to achieve the output
is hidden. Still, the ability of neural network models to handle
non-linear data make such models attractive for modeling financial
instruments.
As another example, chaotic and fractal models may also be
attractive for their ability to handle non-linear data, but errors
in initial values of these models may lead to exponential
divergence over time. Thus, chaotic and fractal models of the
financial instrument may require monitoring and occasional
re-initialization to maintain their accuracy. This
re-initialization procedure typically does not pose a problem in
modeling many financial instruments where a stream of new input
data may be available to update the initialization almost
continuously.
Other considerations for the technical analysis investor include
the type of data to apply to the model and the type of data to be
determined by the model. Many financial models have been created
that utilize historic trade price data as input data.
Trade price data has seemed an obvious choice for the input data of
a financial model. This data includes the actual prices paid for
the financial instrument previously. Additionally, the amount of
data is relatively tractable. As described above, NASDAQ Level II
data for a stock may be refreshed as often as ten times a second,
but the majority of these updates are due to changes in the bid and
ask prices of the stock. Actual trades of the stock occur much less
frequently. Thus, new trade prices are posted at a significantly
lower rate.
Unfortunately, some trades take place at prices that are much
higher or lower than might be expected. These unusual trade prices
may be due to trade-specific concerns, such as gaining a
controlling share in a company, tax issues, etc., that are not
included in the model. If not removed from the input data, these
unusual trade prices may significantly affect the accuracy of the
financial model, but such outlying trades may be difficult to
identify, particularly in a low trade volume financial instrument.
Additionally, depending on the financial market, the trade prices
may not be posted for period time after the trade occurs. Even in
the tightly regulated stock market, there is typically a one to ten
second delay before a trade price is posted. Depending on the
volatility of the financial instrument, the effect of this delay of
the technical analysis investor may range from a minimal issue to
an extreme detriment.
In the exemplary embodiments of the present invention, the actual
series of bid and/or ask prices of the financial instrument are
among the data applied to the model. The bid and ask prices
(quotes) provide a number of advantages as input data for a
financial model. Unlike trade prices, which represent unique
events, bid and ask prices represent a continuous, ongoing record
of the market. At any given time, each financial instrument has
one, and only one, current best bid price and one, and only one,
current best ask price. These quotes remain valid until replaced by
a new quote, at which time the old quote is no longer valid. Thus,
because new quotes do not become effective until posted, access to
new bid and ask prices is nearly instantaneous. This may allow a
financial model based on streams of quotes to operate substantially
in real time.
In a low trade volume financial instrument, the time between trades
may be days or even longer, potentially leading to large
fluctuations between consecutive trade prices and correspondingly
large uncertainty for a technical analysis investor. The quotes may
move numerous times between trades as potential buyers and sellers
seek an appropriate deal, which allows exemplary embodiments of the
present invention to more accurately predict investment significant
parameters of financial instruments. Investment significant
parameters may include any predictions that are desirable to
formulate an investment strategy, for example: future trade prices;
future bid prices; future ask prices; future spreads; fair market
values (FMV) now and/or in the future; expected profit; changes in
prices, spreads, FMV's, or profit between two times; rates of
change in prices, spreads, FMV's, or profit; winners and losers;
and buy/sell instructions. The same advantages may exist for high
trade volume financial instruments as well, only on a different
time scale.
Additionally, the bid prices and the ask prices of a given
financial instrument tend to track one another, thereby maintaining
a relatively constant spread for the financial instrument, as
described above for stocks. Therefore using both the bid and ask
prices in a model may simplify the identification of outlying bid
and ask prices. For example, a sudden jump in the ask price which
is not followed by a corresponding increase in the bid price to
take advantage of the increased price the market is willing to pay
is likely to indicate a misquote or other error. However, if the
increased ask price is real, the bid price is likely to quickly
follow. Thus, exemplary embodiments of the present invention also
may allow technical analysis investors to track sudden changes in a
financial instrument with greater certainty.
FIG. 1 illustrates an exemplary method for modeling an investment
significant parameter of a financial instrument, using a computer,
according to the present invention. The financial instrument may
desirably be a publicly traded financial instrument, but this is
not necessary. It may be any type of financial instrument
including: a stock; a bond; a commodity; a currency; an equity; a
derivative security, or a future.
At least one series of historical bid prices of the financial
instrument or historical ask prices of the financial instrument is
provided as training data to the model, step 100. The provided
series of historical bid and/or ask prices of the financial
instrument may desirably include corresponding bid or ask sizes,
respectively, associated with the quotes. Additionally, the
provided series of historical quotes may desirably be provided as
time series of the historical bid and/or ask prices including
corresponding bid or ask times, respectively. Further a series of
historical spreads between the historical bid prices and the
historical ask prices of the financial instrument may be provided
as well.
It may also be desirable for the historical series of quotes to
include at least one complete consecutive series of: historical bid
and/or ask prices of the financial instrument spanning a
predetermined period of time; or a predetermined number of
historical bid and/or ask prices of the starting from a
predetermined time.
Outlying quotes may desirably be removed from the provided series
of historical quotes, including complete consecutive series, before
the series is (are) applied to the model.
An exemplary method of removing outliers from the provided series
of historical quotes uses the relative stability of the spread
between bid prices and ask price of financial instruments discussed
above. Time series of the historical bid prices of the financial
instrument, including corresponding bid times, and the historical
ask prices of the financial instrument, including corresponding ask
times, are provided. A time series of the spread between the
historical bid and ask prices is calculated from their historical
time series. Outlier bids in the time series of the historical bid
prices may be identified using the time series of the spread, as
may outlier asks in the time series of the historical ask prices.
Quotes that cause the spread to increase, or decrease, beyond
certain thresholds that may vary from financial instrument to
financial instrument, may indicate potential outliers. If the
change in the spread is quickly corrected by a change in the same
quote (e.g. an unusually large jump in the bid price is followed by
a corresponding drop in the bid price), the quote in question may
be identified as an outlier. Conversely, if the change in the
spread is corrected by a change in the other quote (e.g. an
unusually large jump in the bid price is followed by a
corresponding jump in the ask price), the quote in question may
indicate the beginning of a trend up or down in the value of the
financial instrument. Once identified the outliers bid(s) and/or
ask(s) and their corresponding bid or ask times are removed from
the time series.
A financial model type is selected, step 102. As discussed above,
numerous financial model types exist, each with its own advantages
and disadvantages, depending on the financial instrument to be
modeled. One skilled in the art may understand that each financial
model type has at least one variable parameter that may be tuned to
model the behavior of a particular financial instrument. The
variable parameter(s) of the selected financial model type are
initialized, step 104. This initialization may be based on a priori
knowledge of the financial instrument and/or the selected financial
model type, initial values of the series of historical quotes
provided in step 100, a predetermined initial setting, or a
combination of these methods.
The series of historical bid prices and/or historical ask prices
are applied to the initialized financial model type as training
data to estimate the variable parameter(s), step 106. In an
exemplary embodiment of the present invention, the financial model
type selected in step 102 is used to calculate at least one of a
predicted bid price or a predicted ask price of the financial
instrument. This calculation may be based on a set of historical
quotes that includes a predetermined quantity of consecutive
historical quotes from the series of historical bid prices and/or
historical ask prices provided in step 100. The calculation may be
repeated for each set of historical quotes to calculate a plurality
of predicted bid prices and/or a plurality of predicted ask prices.
These predicted quotes may then be compared to the series of
historical quotes provided in step 100. The variable parameter of
the selected financial model type is varied based on the
differences between the predicted quotes and the series of
historical quotes. The calculations and comparisons may be repeated
until the variable parameter has been estimated and the behavior of
the financial model matches the historical behavior of the
financial instrument to within a predetermined degree of accuracy,
i.e. the predicted quotes substantially correspond to the series of
historical quote.
It is noted that the use of the training data provided in step 100
may depend on the type of financial model selected in step 102. For
example, in a non-linear analytic model differences between the
output of the model and the series of historical quotes may be used
as feedback in an estimation maximization algorithm or other
recursive algorithm to adjust the model parameters. In another
example, time series of historical quote prices and corresponding
quote times may be used as training data for a time delay neural
network model of the financial instrument.
It is noted that other training data which may be provided in step
100, such as: a time series of the spread between the historical
bid prices and the historical ask prices; a series of historical
trade prices; and/or extrinsic data, including market indices or
related financial instruments, may also be applied to the selected
financial model type to improve the estimation of the variable
parameter(s).
FIG. 2 illustrates an exemplary method for predicting an investment
significant parameter of a financial instrument, using a computer,
according to the present invention. A model of the bid prices
and/or the ask prices of a financial instrument, based on a set of
historical quotes of the bid and/or ask prices, is provided, step
200. This model may be desirably generated using an exemplary
method of FIG. 1 described above.
At least one of a bid stream of the bid prices of the financial
instrument or an ask stream of the ask prices of the financial
instrument is selected, step 202. This selection may desirably be
based on whether the bid prices, the ask prices, or both of the
financial instrument are modeled by the model provided in step 200.
Although not necessary, for many financial instruments,
particularly high trade volume financial instruments, it may be
desirable for the quote stream(s) selected in step 202 to be a real
time bid stream of the bid prices of the financial instrument
and/or a real time ask stream of the ask prices of the financial
instrument. This may allow for the model to predict, substantially
in real time, at least one of future bid prices of the financial
instrument or future ask prices of the financial instrument,
allowing a technical analysis investor using an exemplary
embodiment of the present invention to react substantially faster
to changes in the financial market. Such improved reaction time may
greatly increase the potential for profits by such an investor.
Desirably, the selected quote stream(s) may include the same
variables as the series of historical data used for generating the
model provided in step 200. For example, if a historical time
series of the bid prices of the financial instrument, including
corresponding bid times is used during generation of the model,
then it is desirable for the quote stream selected in step 202 to
be a bid stream of the financial instrument, including the bid
prices and corresponding bid times. As another example, if a
historical series of the ask prices of the financial instrument,
including corresponding ask sizes is used during generation of the
model, then it is desirable for the quote stream selected in step
202 to be an ask stream of the financial instrument, including the
ask prices and corresponding ask sizes. Additionally, if other data
is included with the series of historical data used for generating
the model provided in step 200, such as: a time series of the
spread between the historical bid prices and the historical ask
prices; a series of historical trade prices; and/or extrinsic data,
it is desirable for those types of data to be included with the
quote stream(s) selected in step 202.
The selected quote stream(s) is (are) applied to the model, step
204. As discussed above with reference to the series of historical
quotes used in the exemplary method of FIG. 1, it may be desirable
to identify and remove outliers in the selected quote stream(s). If
both bid and ask streams of the financial instrument, including
corresponding bid and ask times, respectively, are selected in step
202, then a spread stream may be calculated and used to remove
outlying quotes from the bid and ask streams before they are
applied to the model. Such an exemplary method of identifying and
removing outlying quotes is described above with reference to FIG.
1.
The model is operated on the selected quote steam(s) to predict at
least one investment significant parameter of the financial
instrument, step 206. If the selected quote stream(s) is (are) real
time quote stream(s), then the model may desirably be operated to
predict, substantially in real time, the desired future quotes,
trade prices, FMV, or other stream of investment significant
information about the financial instrument.
Whether these predictions are made substantially in real time or
not, it may also be desirable for each of the predicted investment
significant parameters to include a predicted time, for example
future quotes of the financial instrument may desirably include a
predicted quote time (i.e. a bid time or an ask time) associated
with the predicted future quote. Desirably, the predicted time of
each future investment significant parameter prediction may be
within a predetermined period of time after prediction. If the
predicted time is too close to the time that it is predicted, an
investor using the exemplary method may not be able to use the
information before it becomes stale, or if the predicted time is
too remote, then it may be desirable for the model to hold the
prediction until closer to the predicted time, in case new
information arrives that may affect the prediction.
It is noted that such time predictions may not be available if the
selected quote stream(s) only include(s) sequential quotes of the
financial instrument without corresponding quote times. In this
situation, the predicted future investment significant parameter(s)
of the financial instrument may represent the next anticipated
value of the trade, bid, and/or ask price of the financial
instrument or a short-term buy/sell instruction, etc. For many
technical analysis investors, this information may be adequate and
removing the additional temporal variables may significantly
simplify the model.
Another feature that may be desirable is a confidence level
associated with each of the predicted investment significant
parameter. Many financial model types include calculation of such
confidence level. Thus, the model may desirably predict one or more
future investment significant parameter of the financial instrument
that includes a corresponding confidence level.
The model provided in step 200 may also be dynamically updated
using the selected quote stream(s) as additional training data. The
quotes of the quote streams may be used as additional historic
quotes to continually refine the estimate(s) of the variable
parameter(s) of the model using the exemplary method of FIG. 1.
FIG. 3 illustrates an exemplary method for performing automated
trades of at least one financial instrument, using a computer,
according to the present invention. The financial instrument(s) may
be selected from a single type of financial instrument such as
publicly traded stocks, for example, or may include a number of
financial instruments selected from one or more types of financial
instruments, such as stocks, bonds, commodities, currencies,
equities, derivatives, and futures.
At least one investment significant parameter of each financial
instrument is predicted, step 300. The predicted investment
significant parameter(s) of each financial instrument are predicted
using at least one of a bid stream of the bid prices of the
corresponding financial instrument or an ask stream of the ask
prices of the corresponding financial instrument. Desirably, the
predicted investment significant parameter(s) may be determined
using one or more of the exemplary methods of FIGS. 1 and 2 as
described above. The quote stream(s) used may include real time
quote streams of each financial instrument or may include some
quote streams that are not provided in real time. As described
above the predictions may utilize additional data, such as: spread
streams for one or more of the financial instruments; trade price
streams for one or more of the financial instruments; and/or
extrinsic data, including market indices or related financial
instruments.
Additional prediction information such as predicted times and
corresponding confidence levels of each of the one or more
corresponding investment significant parameter may be desirably
provided as well for one or more of the financial instruments.
If the investment significant parameter(s) predicted in step 300
are numerical values, then, for each of the financial instruments,
the predicted investment significant parameter(s) are compared to
at least one of the most recent corresponding value(s) of the
corresponding financial instrument to determine trend data of the
corresponding financial instrument, step 302. If the predicted
investment significant parameter(s) for a financial instrument
include corresponding parameter confidence level(s), the trend data
determined for the financial instrument may desirably include a
corresponding trend confidence level based on the corresponding
parameter confidence level. The buy criterion and the sell
criterion for a given financial instrument may be varied based on
its trend confidence level. Alternatively (or additionally), the
trend confidence level(s) of the financial instrument(s) may be
tracked to help determine when to place buy and/or sell orders.
If more than one financial instrument is being traded using the
exemplary method of FIG. 3, then an exemplary joint trade strategy
approach may be used to determine the trend data corresponding to
each of the financial instruments in step 302. In this exemplary
approach, for each of the financial instruments, the predicted
investment significant parameter(s) are compared to at least one
most recent corresponding value of the corresponding financial
instrument to determine a predicted change in the predicted
parameter of the corresponding financial instrument. The plurality
of predicted changes in the predicted parameters determined may be
analyzed to formulate a joint trade strategy for the financial
instruments over a predetermined period time to maximize
anticipated return. The trend data corresponding to each of the
financial instruments may then be set based on the joint trade
strategy.
One skilled in the art may understand that various algorithms may
be used to analyze the plurality of predicted changes and formulate
the joint trade strategy including: expert system models; linear
analytic models; non-linear analytic models; chaotic models; neural
network models; time delay neural network models; Markov-chain
Monte Carlo models; wavelet transformation models; regression
models; fractal models; support vector machine models; or Bayesian
models.
The corresponding trend data of each financial instrument is
compared to the buy criterion of that financial instrument, step
304. If the corresponding trend data of a financial instrument does
not meet the corresponding buy criterion, then the trend data is
compared to the sell criterion of that financial instrument, step
308. As described above the buy and/or sell criteria may be varied
based on a trend confidence level of the financial instrument.
Alternatively, these criteria may be predetermined, based on a
priori knowledge of the financial instrument or financial market,
or may be determined by the corresponding model of the financial
instrument. Additional factors may also affect these criteria. For
example, the amount of cash, or credit, available to purchase
financial instruments may affect the buy criteria of the financial
instrument and whether any quantity of a given financial instrument
is owned may affect the sell criteria of that financial
instrument.
It is noted that the order of steps 304 and 308 may be reversed, or
these steps may be performed substantially simultaneously without
departing from the present invention.
If the buy criterion is met in step 304 for a given financial
instrument, then a buy order for that financial instrument is
automatically placed, step 306. Likewise, if the sell criterion is
met in step 308 for a given financial instrument, then a buy order
for that financial instrument is automatically placed, step 310. If
neither criterion is met, then no order is place for the given
financial instrument, step 312.
When either a buy order is placed in step 306 or a sell order is
placed in step 310, a buy size of the buy order or a sell size of
the sell order may desirably be determined. The buy size may be set
to a predetermined size, a predetermined total price, or may be
based on the amount of cash, credit, and/or other liquid assets
available for the purchase. The sell size may be set to a
predetermined size, a predetermined total price, or may be based on
the quantity of the financial instrument available for sale by the
investor. If a corresponding trend confidence level has been
determined, then the buy size of the buy order, or the sell size of
the sell order, may be determined based on the corresponding trend
confidence level.
Alternatively, a buy size and a sell size for each financial
instrument may be determined based on a comparison of the trend
data and the buy criterion of the financial instrument. A buy order
is then automatically placed for each financial instrument for
which the buy size is greater than zero and a sell order is
automatically placed for each financial instrument for which the
sell size is greater than zero.
If the investment significant parameter(s) predicted in step 300
are non-numerical values, e.g. buy/sell instructions and/or winner
and loser predictions, then the predicted investment significant
parameter(s) may be used directly to determine what trade orders
should be placed. If the predicted investment significant
parameter(s) for a financial instrument include corresponding
parameter confidence level(s), then these parameter confidence
level(s) may be used to determine the size of any trade orders that
are to be placed. In the case where more than one financial
instrument is being traded, the parameter confidence level(s) may
be used to develop a joint trade strategy among the financial
instruments that may include determining the size of any trade
orders to be placed.
It is noted that both numerical and non-numerical investment
significant parameters may be predicted by the exemplary model in
step 300 of FIG. 3. In this situation, the investment significant
parameters may be used in combination to determine a (joint) trade
strategy. In determining a trade strategy in this manner, each
investment significant parameter being given a weight in deciding
whether to place buy or sell orders. Alternatively, different
investment significant parameters may be used for different
determinations, e.g. a buy/sell instruction may be used to
determine whether to place an order to buy or sell the
corresponding financial instrument and the predicted profit of each
trade may be used to determine order sizes between several
financial instruments.
The various exemplary embodiment of the present invention may be
carried out through the use of a general-purpose computer system
programmed to perform the steps of the exemplary methods described
above with reference to FIGS. 1, 2, and 3. Exemplary
general-purpose computer systems may include personal computers,
work stations, distributed processing computer networks, and
parallel processing computer systems. Parallel or distributed
processing may be desirable for substantially real time
applications involving the substantially concurrent prediction of
future quotes for a plurality of financial instruments. Dedicated
special-purpose computing systems may also be designed for
performing exemplary methods of the present invention as well.
Additionally, it is contemplated that the methods previously
described may be carried out within a general purpose computer
system instructed to perform these functions by means of a
computer-readable medium. Such computer-readable media include;
integrated circuits, magnetic and optical storage media, as well as
audio-frequency, radio frequency, and optical carrier waves.
Although the invention is illustrated and described herein with
reference to specific embodiments, the invention is not intended to
be limited to the details shown. Rather, various modifications may
be made in the details within the scope and range of equivalents of
the claims and without departing from the invention.
* * * * *